Adaptive scale target tracking method based on structured support vector machine

Abstract

The invention discloses an adaptive scale target tracking method based on a structured support vector machine, which mainly solves the problems of large amount of calculation in the adaptive scale and tracking in the prior video target tracking based on the structured support vector machine. The implementation steps are as follows: firstly establish a structured output support vector machine model, and add scale variables to the output of the model; then update the decision function by using the image frame that has determined the target; finally, decompose the target tracking into rough tracking and fine tracking. Tracking estimates the target position from a small number of candidate samples to narrow the target search range, and then determines the position and scale of the target on the basis of rough tracking through fine tracking. The invention realizes the self-adaptive scale target tracking, reduces the calculation amount in the tracking process, improves the real-time effect, and can be used to determine the precise position and real-time scale of the target in video monitoring.

Description

technical field [0001] The invention belongs to the technical field of image processing, and in particular relates to a video target tracking method, which can be used to realize precise tracking of targets. Background technique [0002] Automatic target tracking based on image and video sequences is an important content in the field of image processing and pattern recognition, and has been widely used in industry, transportation and other fields. The tracking model established in tracking still cannot completely overcome the problems of light intensity change, background change, occlusion, robustness and so on. [0003] In the paper "Struck: Structured Output Tracking with Kernels" (IEEE International Conference on Computer Vision, 2011, 263-270), Sam Hare et al. proposed a sample structured output support vector machine for video sequence target tracking. The method first initializes the classifier with the first frame image, then directly uses the classifier to predict t...

AI Technical Summary

Problems solved by technology

This method has achieved good results in terms of occlusion and robustness; however, this method still has the following shortcomings: First, it cannot achieve adaptive scale tracking in video tracking, that is, when the target is far away from the camera and close to the camera, the tracking frame It cannot be adjusted adaptively; the second is that rough position estimation is not performed when predicting the target position, resulting in an excessively large search range, which has a great impact on storage and calculation, and affects real-time performance

Images